DocFetcher vs Precision File Search

DocFetcher is a free, open-source desktop application for searching the contents of files stored on a computer. It presents selected results with a text preview and highlighted matches, and search results can be narrowed by file size, type and location. Its query syntax supports AND, OR and NOT, wildcards, phrase and fuzzy searches, proximity search and boosting. Supported content includes Microsoft Office files, Outlook PST, OpenOffice, PDF, EPUB, HTML, plain text, RTF, media metadata, Visio and SVG. It can search ZIP, 7z, RAR and tar archives, including archives nested inside other archives. Users create indexes for folders; the application detects changes while running, and a background daemon records changes while it is closed for processing at the next launch. Portable editions keep indexes and settings in the application folder and can be moved between computers. The maker says the application collects no private data and indexes nothing by default. It runs on Windows, macOS and Linux. The project is legacy software: no new features are planned, though bug fixes and support continue.

Precision File Search is a free desktop application for finding, classifying, and understanding local files on Windows, macOS, and Linux. It combines classic search with semantic search using a retrieval-augmented generation pipeline, and can answer natural-language questions with summarized responses. A trainable machine-learning model sorts scattered documents into categories users define. The maker says the application and AI models can operate locally, without cloud dependency, and that files, queries, and personal data are not sent to external servers. The project lists support for 18 languages, including English, Arabic, Chinese, French, German, Hindi, Japanese, and Spanish. An internet connection is needed on first run to download multilingual AI models; later runs can be offline. PFS supports OpenAI-compatible inference URLs, including online providers and local engines such as Ollama and LM Studio. The README lists 8 GB of RAM for the core platform and a dedicated GPU, with at least 4 GB of VRAM for smaller models, for local LLM use. The project is licensed under MPL-2.0.